IRC logs for #baserock for Wednesday, 2014-10-08

*** thecorconian [~thecorcon@] has quit [Remote host closed the connection]00:37
*** vmeson [~quassel@] has joined #baserock01:31
paulsherwoodso i appear to be in a strange universe where localhost !=
paulsherwoodPING ( 56 data bytes02:09
paulsherwood64 bytes from seq=0 ttl=64 time=0.091 ms02:09
paulsherwoodping localhost02:09
paulsherwoodPING localhost ( 56 data bytes02:09
paulsherwood--- localhost ping statistics ---02:09
paulsherwood6 packets transmitted, 0 packets received, 100% packet loss02:09
*** ratmice_______ [] has quit [Ping timeout: 246 seconds]06:35
*** ratmice_______ [] has joined #baserock06:35
*** mSher [] has joined #baserock06:44
*** dutch [] has joined #baserock07:39
*** tiagogomes [] has joined #baserock07:53
*** tiagogomes [] has quit [Ping timeout: 250 seconds]08:03
*** tiagogomes [] has joined #baserock08:15
Kinnisonpaulsherwood: that happens when people who shouldn't be allowed to operate computers put them on networks08:17
Kinnisonsomewhere a computer *called* localhost will be DHCPing08:18
pedroalvarezheheh, you where trying  to upgrade someone's computer?08:30
*** jonathanmaw [] has joined #baserock08:31
*** ssam2 [] has joined #baserock08:31
straycatfranred, You could add one as a submodule to the other?08:44
pedroalvarezmercurial :S08:46
franredstraycat, that implies modify the repository after lorrying them - I rather copy the content of one in the other when B has to compile08:46
pedroalvarezeverything is on git08:46
franredI did this and it works08:46
KinnisonWe have another case for multiple-repo sources08:47
straycatfranred, sure08:47
franredstraycat, although I will try the submodule approach08:49
pedroalvarezfranred: an example
KinnisonThat is not sufficient08:50
Kinnisonyou need both the .gitmodules *and* the submodule link08:50
Kinnisonuse the git submodule commands08:50
Kinnisondon't try and write it yourself08:50
franredAFAIK, if one repo has submodules, if that repo wants to use the latest version of the submodule it has to update it and push them, how baserock manages this?08:53
KinnisonCurrently a human will have to do it08:54
KinnisonFrankly I doubt firehose ought to know how to08:54
Kinnisonsince normally the submodules will be set by upstream08:54
franredso we need to update 3 parts, the repo to update the submodule, and the references on the chunks?08:54
Kinnisone.g. we wouldn't want to move cgit's git submodule08:54
Kinnisonfranred: yes, you'd need to update the submodule, commit the update and push it (to the enclosing repo) and then update the chunk's entry in the stratum08:55
* Kinnison thinks this is another good indication that we need multi-source chunks08:55
* franred doesn't like submodules at all08:55
*** zarazaimeche [] has joined #baserock08:58
zarazaimeche is now known as Zara08:59
* pedroalvarez waves zarazaimeche 08:59
Zarahi, pedroalvarez :)08:59
ratmice_______Kinnison: done some of that with iirc git read-tree checking out multiple repos into a working dir, then frobbing the GIT_DIR env var based on which which repo i want git commands to work on, can be interesting when the repos overlap/checkout order matters09:18
KinnisonFrankly we probably don't need that per-se09:19
KinnisonJust the ability to effectively provide a "repo+sha+location" set09:19
Kinnisonwhere the base repo+ref is what goes into
Kinnisonand everything else is relative to (and locked inside) that dir09:19
pedroalvarezthat makes sense to me09:20
ratmice_______yeah, it was a fairly weird case configure script from one repo would test for dirs in the top-level srcdir from another repo09:21
Kinnisonratmice_______: sounds uncomfortable09:21
* Kinnison wonders why ratmice_______'s tail has grown quite so long :-)09:21
* ratmice_______ too09:22
ratmice_______I actually just want to discover what happens when it hits some arbitrary limit09:23
ratmice_______anyhow in particular it was gcc+binutils+newlib repos stitching together the old 'uberbaum' tree out of git repos09:24
KinnisonMmm, I think I remember having to do something similar waaay back in about 2004 or so09:24
ratmice_______it actually worked fairly well, since you could easily see divergence between files existing in all the repos09:24
KinnisonBack on MIPS09:24
KinnisonAlthough I think that was glibc not newlib09:25
ratmice_______hmm, not aware of glibc ever being buildable like that09:25
KinnisonIt was some odd thing that MIPS did internally09:25
KinnisonI think it was more that we layered things and patched them09:25
* Kinnison wasn't meant to be on toolchains though so wasn't deeply involved beyond writing build tools09:25
ratmice_______anyhow, i liked it, really cut down on # rebuilds of all those various libs09:26
pedroalvarezgreat, I finally confirmed the error that the jetson board was having to boot the latest relase09:31
paulsherwoodKinnison: nonetheless, given that localhost can be screwed, what about my patch?09:32
paulsherwoodpedroalvarez: what was it?09:32
richard_mawpaulsherwood: I just sent you a reply about your patch09:32
pedroalvarezpaulsherwood: cma=256M in KERNEL_ARGS09:32
pedroalvarezpaulsherwood: after removing it, everything worked09:33
richard_mawI have no problem with your patch, but it includes a command to fix it globally, and I'm going to try out the gerrit workflow when submitting the config fix09:33
pedroalvarezrichard_maw: :)09:33
paulsherwoodrichard_maw: ok, thanks09:34
pedroalvarezso, my question regarding fixing  the release. What should I do?09:34
paulsherwood14.40.1 09:34
pedroalvarezPut the patches on the top of the release branch?09:34
pedroalvarezand tag it as paulsherwood suggested?09:35
Kinnisonworks for me09:35
ssam2pedroalvarez: out of curiousity, how to did you work out that was the problem ?09:38
pedroalvarezflashing, and reflashing and reflashing09:38
pedroalvarezwith various conbinations of rootfs and u-boot, I saw it wasn't u-boot the problem09:38
pedroalvarezand it was something in the rootfs09:39
pedroalvarezThe only thing in the rootfs that could be causing problems was extlinux.conf09:39
pedroalvarezthen I remembered one change in the KERNEL_ARGS that radiofree did for the genivi system09:40
pedroalvarezso, I wondered: Is that the problem? 09:40
pedroalvarezand it was09:40
pedroalvarezthe slowest part of this proccess has been: reflashing the jetson09:40
pedroalvarezI know there is a way to mount the disk of the jetson and modify things.09:41
paulsherwoodpedroalvarez: why were you not using the cycle approach? (i'm sure there'll be a reason)09:41
Kinnisonpaulsherwood: presumably because it wasn't booting09:41
Kinnisonpedroalvarez: In theory you can run some uboot command which exposes the storage as a USB disk09:42
pedroalvarezand at the beggining I thought the problem was on u-boot, and to reinstall it you have to flash it09:42
paulsherwoodpedroalvarez: ridght09:42
pedroalvareznow I wonder if reverting the upgrade of u-boot makes sense, but that means reflash the jetson  to test it09:43
* franred is not good typing in the correct window....09:43
Kinnisonpedroalvarez: the uboot update was needed for ethernet and sata no?09:43
pedroalvarezI shall try then09:44
pedroalvarezthe only thing I know, is that James thought that the u-boot upgrade was causing problems09:46
KinnisonI think it might be truncating the kernel args if they're too long09:49
KinnisonI heard him mumble something about that last week09:49
paulsherwoodpedroalvarez: i have the old u-boot on a jetson09:50
paulsherwoodbut actually, given my localhost nonsense, i can't test for you09:51
Kinnisonpaulsherwood: fix your /etc/nsswitch.conf to put 'files' before the rest of the entries on the hosts line09:52
Kinnisonpaulsherwood: that should cause it to ignore the network for localhost09:52
paulsherwoodiin the meantime i got this
Kinnisonassuming I understood richard_maw's mail properly09:52
richard_mawpedroalvarez: so before I can push to the gerrit, I need to register an account?09:53
Kinnisonssam2: Any idea what could cause paulsherwood's problem? ^^^09:53
richard_mawKinnison: no, you need to my myhostname just after files09:53
richard_mawrather than at the end09:53
Kinnisonrichard_maw: oh right09:53
paulsherwoodKinnison: it already is: hosts: files mdns4_minimal [NOTFOUND=return] dns mdns4 myhostname09:53
* Kinnison misunderstood your seddery09:53
Kinnisonpaulsherwood: so move myhostname to between files and mdns4_minimal09:53
pedroalvarezrichard_maw: I'm not sure, but I think that makes sense that you have an user to be able to send a patch09:54
richard_mawpaulsherwood: are you missing VERSION_LABEL in that deploy you just fpasted?09:54
paulsherwoodoh, probably.09:55
paulsherwoodcan we make deply idiotproof? :)09:55
Kinnisonthe world will invent better idiots if we try09:55
pedroalvarezwe should check that in the .check extension09:55
Kinnisonwe should mandte VERSION_LABEL if you're using ssh+rsync though09:55
paulsherwoodwhat about making it default to TEST?09:56
KinnisonI think if you're applying an update you should be specifying the basis (default to FACTORY) and the target (mandatory)09:56
Kinnisonand the .check can fail quickly09:57
pedroalvarezsounds like updating .check is the right solution09:58
richard_mawpedroalvarez: :-( the only OpenID I have that is listed is Google, and Google no worky09:58
paulsherwoodi note your suggestion. agree with default to factory (not FACTORY) but see no problem with defaulting to TEST09:58
pedroalvarezrichard_maw: google not worky, and they don't care about that09:58
pedroalvarezrichard_maw: you should have other openid providers09:59
pedroalvarezI used launchpad yesterday (whch uses ubuntu one :( )09:59
pedroalvarezalso yahoo works fine09:59
Kinnisonrichard_maw: I can recommend stackoverflow's openid provider09:59
Kinnisonit works and is easy to use09:59
paulsherwoodmwilliams_ct: aren't you working on an open_id server?09:59
* Kinnison also recommends you set it up as a delegate for your real website10:00
* Kinnison 's OpenID is
KinnisonJust needs a couple of <link> tags on your front page10:00
richard_maweww, the registration page doesn't label what each text box is for10:01
Kinnisonit's javascripted10:02
Kinnisonemail, realname, password, password-again, captcha, respectively10:02
ssam2Kinnison: how do I use stackoverflow as an OpenID ?10:02
ssam2I have
Kinnisonssam2: look at openid.stackexchange.com10:02
ssam2I get 'Provider is not supported, or was incorrectly entered.'10:03
Kinnisonssam2: and the <link> tags on
ssam2no sign of our Gerrit instance on that log10:03
Kinnison(actually I should try the gerrit with this)10:03
Kinnisonpedroalvarez: can you provide me with a link to the gerrit, I've lost it10:03
pedroalvareznote: Gerrit doesn't have configured the email, so you can't do things like change your email address10:04
franredI couldn't manage to have and stackexchange account loggin into gerrit10:04
* richard_maw gets verification failed10:04
KinnisonLooks like it works with the approach I take10:04
KinnisonI have delegated openid via my personal website10:04
ssam2I have to use the link from
KinnisonBetter is to use something which delegates the ID10:05
ssam2to delegate that from my personal website do I just need to add the two <link> tags to index.html ?10:05
Kinnisonyou can crib the approach from
ssam2that's a bit much to ask every Baserock contributor to do, however :)10:06
KinnisonI'm surprised the google login thing didn't work for richard_maw10:07
Kinnisonaah it's 'cause google deprecated their openid provider10:07
Kinnison*plus* we're presenting an IP address not a domain name10:08
Kinnisonwhich they don't like10:08
KinnisonHowever, if we could have a basic provider then it'd be a good thing10:08
ssam2yeah. Seems like there's no way to get a StackExchange OpenID to work for me10:10
ssam2I get forwarded to StackExchange and back, but I never actually get logged into Gerrit10:10
ssam2a problem in our instance franred thinks10:10
KinnisonI can log in with my delegated ID10:11
straycatCan tarballs be lorried from multiple urls?10:13
straycatThe spec seems to have a single url field, so I'm guessing we can't.10:16
KinnisonNo it's only one url10:17
pedroalvarezstraycat: does it makes sense? I mean, what use case are thinking of10:17
pedroalvarezs/are/are you/10:18
straycatI was thinking it might be simpler to lorry all versions of a package so that you can easily use/switch between serveral versions of a package.10:19
KinnisonLorry one, then update the url, rinse and repeat10:20
Kinnisonit the current workflow for that10:20
* straycat nods10:21
pedroalvarezoh, I see why you are looking at this10:21
richard_mawpedroalvarez: which url should I push to?10:21
pedroalvarezis a definitions patch?10:21
richard_mawah, there's tiny buttons10:21
richard_mawnevermind, got it10:21
pedroalvarezrichard_maw: the url is also in the cgit interface10:21
pedroalvarezport :8010:22
richard_mawit's a different URL for submitting changes isn't it?10:22
richard_mawbah! it's refusing my commits, because it doesn't like that I don't have an e-mail registered, because I get internal server errors10:23
pedroalvarezthe email in the commit has to be the same email that you used to register10:24
richard_mawit didn't let me enter an e-mail at registration time10:24
pedroalvarezknown error, not email server configured10:25
pedroalvarezmaybe worth fixing that10:25
richard_mawhow did you manage to have an e-mail registered anyway?10:25
pedroalvarezI'm trying to figure it out10:25
paulsherwood2014-10-08 09:46:06 [Build 334/349] [enlightenment] Cloning upstream:enlightenment/enlightenment10:26
paulsherwood2014-10-08 10:25:19 [Build 334/349] [enlightenment] Creating staging area10:26
Kinnisonpaulsherwood: That's an uncomfortbly long clone time10:26
pedroalvarezyou are far away from g.b.o 10:26
* jmacs notes that python's platform.linux_distribution returns empty strings10:26
pedroalvarezrichard_maw: seems like the launchpad openid provides the email when registering10:27
richard_mawI see. Well mine doesn't, which means I can't use Gerrit10:28
pedroalvarezI'd like to sort out the email server config10:29
* Kinnison suggests that exim4 is pretty easy to get going,10:31
richard_mawpedroalvarez: I'd suggest getting it working is a reasonably high priority, since it'll make patch review easier in the long run10:36
pedroalvarezi agree10:37
ssam2jmacs: in Baserock, you mean ?10:38
jmacsssam2: Yes10:44
jmacsI'm just looking to see how it's meant to work10:44
jjardonpaulsherwood: should I assume your changed your -1 to a +1 in the llvm series when I change to build 3.3.1 instead 3.5?10:45
ssam2I'm thinking that the import tooling should move out of its branch of morph.git and into its own repo10:49
ssam2any opinions on this, or names for it?10:49
ssam2if noone complains I think git:// should be its home10:49
KinnisonI'm happy with that10:55
wikicatWiki change:
ssam2ok. Can someone with appropriate permissions create that? I guess that's richard_maw or Kinnison, who have both left for lunch.11:04
ssam2or not.11:05
Kinnisonnow I shall lunch11:08
*** JPohlmann [] has joined #baserock11:10
*** JPohlmann [] has quit [Changing host]11:10
*** JPohlmann [~jannis@xfce/core-developer/JPohlmann] has joined #baserock11:10
jmacsIs "build-essential" the first thing that gets built in a baserock system?11:28
jmacsOK. I was just a bit confused as to why a stratum that seems to be mainly about compilers writes /etc/passwd etc11:36
ssam2from stage 3 of build-essential onwards, everything runs in a chroot, so it needs to be a viable chroot environment11:44
ssam2fhs-dirs is kind of treated as 'everything that doesn't fit in the other parts of bootstrap but is needed for the chroot to work' chunk11:44
jmacsYes, I thought that would be the case11:45
wikicatWiki change:
richard_mawpedroalvarez: how did the distbuild test with the .cache_key fix in worker_build_scheduler go?12:32
jmacsWhat's VERSION_LABEL meant to contain? Is it arbitrary?12:40
richard_mawyep, but I don't think it'll handle / too well12:40
jmacsIs there another environment variable that contains the version of baserock? e.g. "14.40"12:42
Kinnisonjmacs: the version label is used to control the variant of the system which is booted12:42
Kinnisonjmacs: the default label being 'factory'12:42
KinnisonI doubt there's anything with a Baserock version per-se currently12:42
KinnisonWe should probably make our lsb-release stuff work that out12:42
jmacsKinnison: That's what I'm trying to implement. Do we already have lsb-release stuff?12:43
KinnisonI believe we have a small amount of lsb stuff12:43
richard_mawwe've got a minimal /etc/os-release12:43
Kinnisonhow complete / correct it is, I have no idea12:43
richard_mawit basically just says "Baserock"12:43
jmacsThere is no /etc/lsb-release on the system I have here12:44
Kinnisonthen we're lacking all of that12:45
jmacsAre there conflicting standards here? Should we have both lsb-release and os-release?12:45
richard_mawI think the key difference is that lsb-release is executable, and os-release is machine parsable12:46
pedroalvarezrichard_maw: it's working fine12:58
richard_mawis that an endorsement to merge :-)12:59
pedroalvarezrichard_maw: is a +1 to merge if you fix the missing artifact.cache_key13:00
pedroalvarezand I believe you have already done that13:01
ssam2I don't have /etc/lsb-release on Fedora13:02
ssam2I do have /etc/os-release with VERSION and VERSION_ID fields13:02
ssam2VERSION_LABEL is user-defined, I tend to put the deployment date, but it's just a mnemonic for the user / admin of the system13:03
ssam2possibly it should be hidden in future13:03
KinnisonI have a program called lsb_release13:03
Kinnisonwhich is in /usr/bin13:03
ssam2the thing is that Baserock systems don't really have a version beyond 'SHA1 of the repo it was built from'13:04
KinnisonI wouldn't object to a configuration extension which could augment that with a release number when you deploy13:04
KinnisonIf Python can deal with getting that at runtime instead of build time13:04
ssam2that'd be cool, but I worry about potential for confusion with VERSION_LABEL13:05
jmacsAh, I misread It should in fact check os-release13:05
Kinnisonthen we need to know what's missing and how to fix it13:06
jmacsssam2: What's the first line of /etc/os-release on your Fedora system?13:10
jmacsI don't think Python will do anything with /etc/os-release. It reads it because it looks for any file called .*[_-](release|version) in /etc, but it only checks the first line for something of the form "distro release x.y (codename)"13:14
jmacsWhich wouldn't be in os-release as that's only meant to contain assignments13:15
jmacsTo get Python to work we either need an lsb-release or a baserock-release13:15
Kinnisonwill it read *any* file named .*[_-](release|version) ?13:15
jmacsSo far as I can tell, yes13:16
KinnisonDebian's os-release starts PRETTY_NAME="Debian GNU/Linux 7 (wheezy)"13:16
Kinnison(for stable)13:16
Kinnisonperhaps whoever wrote the code for python was a debianite but not very sensible?13:16
jmacsYes, but it also has debian_release, I think13:17
jmacsIt'll check os-release, not find a match, and continue to debian_release13:17
Kinnisondebian_release contains just the version number (7.4) iirc13:18
jmacsIt's debian_version, and yes, just the number13:18
jmacsWhich probably means platform will fall through to other methods of detecting the OS13:19
ssam2FWIW, I see:13:23
ssam2>>> platform.linux_distribution()13:23
ssam2('Fedora', '20', 'Heisenbug')13:23
jmacsI've just added an /etc/lsb-release to my baserock system and I still get ('', '', ''). Oh well.13:24
ssam2richard_maw: I'm about to update definitions.git after merging a morph.git branch, so if you've not done so yet, don't worry about it13:43
richard_mawalready done13:43
richard_mawyou'll get a conflict13:43
richard_mawcurrently sending mails to acknowledge13:44
ssam2ok, i'll pull first13:44
pedroalvarezrichard_maw: email working on gerrit :)13:59
pedroalvarezKinnison: thanks for your help :)14:00
richard_mawit's not errored when trying to register a mail, but it's not turned up in my inbox yet14:01
pedroalvarezrichard_maw: can you check your spam?14:01
richard_mawthere it is!14:02
* pedroalvarez dances14:02
* Kinnison waits for the greylisting on his domain to expire for the registration email14:05
Kinnisonpedroalvarez: the one came through, but my is probably still suffering from greylisting14:07
pedroalvarezlet's see if I can do anything useful with google, greylist and exim414:08
KinnisonOh it's fine14:09
KinnisonI just need to wait the greylisting timer and the retry timer on the gerrit server (which is likely ca. 15m or 30m depending)14:09
pedroalvarezfrom exim4 logs: This domain has greylisting enabled.14:10
rjekThat sounds like an error message from my mail server.14:11
Kinnisonif you run 'exim4 -bpr' then you'll see it sat in the queue for retry later14:11
rjek2014-10-08 15:03:19 H=(gerrit) [] F=<> temporarily rejected RCPT <>: This domain has greylisting enabled. If you receive this email message, it is likely that your mail server is incorrectly configured. Please contact the person you wish to contact using another method.14:11
Kinnisonrjek: aye, it's all expected and good14:11
rjekGood good!14:11
* Kinnison put in his work address which is, IIRC, postfix and no greylisting, and his home address which is exim4 and greylisting14:12
Kinnisonpedro I believe tried gmail14:12
pedroalvarezyup, and it worked14:13
richard_mawhttp:// may be of interest14:14
KinnisonIs it right to hardcode the entire thing rather than to sed it like you suggesed before?14:15
richard_mawI want to drop it from the delta, but we can't make changes to non baserock/ repositories in gerrit currently14:16
richard_mawanyway, if I was sedding, there'd be two different things you need to look up before you know the result, and sed is not easily comprehensible14:17
richard_mawanway, I think we need to get used to the comment buttons, so it may make more sense to do this on the instance14:17
pedroalvarezI've made a comment here: I wonder if the people can see it14:20
richard_mawpedroalvarez: I don't see it14:22
richard_mawbut I do see Kinnison has added himself as a reviewer14:22
richard_mawand he has a review in the history14:23
* Kinnison has reviewed it too14:23
Kinnisonpedroalvarez: I can see your comment in the history of the change14:25
pedroalvarezso it has 2 +1s, but it doesn't have a +2 yet14:26
KinnisonNot sure what the rules are by default in gerrit14:26
richard_maw:-\ Now I want a Zuul instance to do the merge14:27
pedroalvarezzuul needs a JSP like tomcat14:27
* Kinnison 's wrist vibrates14:27
richard_mawpedroalvarez: :-|14:28
Kinnison(it's not an allergic reaction to JSP, but an email notification)14:28
* Kinnison pukes all over pedroalvarez -- that's the reaction to JSP14:28
Kinnison2453 N + Oct 08 Pedro Alvarez   (0.7K) [Gerrit Code Review] Email Verification14:28
pedroalvarezi can be wrong, but that's whatI read14:28
*** genii [~quassel@ubuntu/member/genii] has joined #baserock14:28
Kinnisonpedroalvarez: looks like the MTA on your gerrit instance is doing the right thing with greylisted domains \o/14:29
pedroalvarezKinnison: yay \o/14:29
Kinnisonpedroalvarez: how did you +2 ?14:39
pedroalvarezfrom the admin user14:40
pedroalvarezbut if 1+1 is not 2, then some people has to have rights to +214:40
KinnisonNo, I think the idea is that it has 'n' +1s14:41
KinnisonOverall I'd like to see a rule of ">= n +1s from humans, and a +1 from CI"14:41
KinnisonBut I don't know if we'll ever get there14:41
jjardonpedroalvarez: that is looking cool ;)14:42
* Kinnison shall try and get git review and gerrymander going at some point14:42
Kinnisonunless someone else wants to first?14:42
jjardonI have a question about the troves: why are they needed? My understanding is for reproducibility reason: in case upstream repos dissapears or something. Also to have everything in git. Are those the only reasons?14:46
jmacsGrr. Reading /etc/lsb-release is a Ubuntu-specific patch to python.14:46
Kinnisonjmacs: bleh14:47
Kinnisonjmacs: write a baserock-release with the data in the right format to shut python up/14:47
ssam2jjardon: also for speed. In some use cases it's good to have source code mirrored in a machine on the local network14:47
richard_mawjjardon: also, to avoid overloading upstream's repository14:47
ssam2jjardon: although Morph also does local caching of Git repos, so there are currently two components doing caching which is not ideal14:48
richard_mawstick a trove on devel machines!14:48
jmacsKinnison: I can do that, but I'll still have to patch Python since it won't recognise /etc/baserock-release unless I add baserock to its approved list of distributions (as I've now discovered)14:48
Kinnisonprobably best not to need a python patch14:49
* Kinnison wonders if there's anything we can do instead14:49
richard_mawpropose a proper /etc/os-release parser, so there doesn't need to be distro-specific hacks anywhere?14:50
jmacsPresumably in baserock there's some way to patch the copy of python we produce14:50
jmacsrichard_maw: That sounds sensible14:50
richard_mawyeah, but we try to avoid unnecessary deltas if possible14:50
richard_mawif we do need a delta, then we need it to be something that's upstreamable14:51
richard_mawI just wonder whether it would count as a bugfix, so python 2.7.x would take it14:51
jjardonssam2: richard_maw Id say those are not really important nowadays: connection speeds are fast! and do not think have overloading problems. Well I meant not important enough to have to lorry every single piece of software you need14:51
jjardonwhat happen if a force-push is done in one of repo? does trove do something spacial or it simply mirror what its in the upstream repo?14:52
straycatTrove also gives you a way to convert non-git to git, and somewhere to put it.14:53
Kinnisonjjardon: By default we whinge about force-pushes and a human has to check on things and update the lorry specs14:53
straycattrove's rulesets won't allow force pushes to certain refs.14:53
richard_mawwhich gives us time to make a copy of the branch if we need to14:53
jjardonok, good. I think we can assume that the content in our trove is free of "missing sha's" problems then. Why then is not allow to use tags as a ref: instead the sha?14:56
ssam2jjardon: "connection speeds are fast"! call me when you're in a hotel.14:56
ssam2or on-site in Eastern Europe14:56
straycatjjardon, you can use the ref if you want, it's better for definitions on gbo to use sha to stop things from changing under peoples' feet.14:57
jjardonstraycat: AFAIK you cant change a tag if you dont force-push?14:58
KinnisonYou can14:58
Kinnisondelete it14:58
Kinnisonwait a bit14:58
Kinnisonupload a new one14:58
KinnisonHell, you can delete tags14:58
jjardonssam2: are you suggesting I should have a trove in my laptop? :)14:59
KinnisonAnd upstream *have* in the past moved tags14:59
Kinnisonesp. ones which are not annotated14:59
KinnisonAlthough I've seen annotated tags being moved too14:59
ssam2jjardon: I'd like that to be possible, yeah15:01
ssam2main problem is it needs to be capable of having a subset of the full content15:02
jjardonKinnison: yeah, but my point is, you cant move tags without force push?15:02
straycatssam2, I don't think running a trove on a laptop would be much fun, given the resource requirements for trove.15:02
ssam2agreed, but I'd like the resource requirements to be lower15:02
* straycat nods15:02
Kinnisonjjardon: You can by deleting it, waiting, and then pushing a new one15:02
Kinnisonjjardon: just like you can with any branch15:02
Kinnisontags are just refs15:03
pedroalvarezKinnison: I believe that you can now +2 patches15:04
pedroalvarez(in definitions.git)15:05
KinnisonI'll become drunk with powah15:05
franredyou can do +2 but I think you are not able to merge the patch ... yet :P15:06
pedroalvarezi think that now he is15:06
franredummm so much powah!! Kinnison is becoming Spiderman15:07
rjekThingy Garfield or Toby Whatisface?15:08
straycatI take it we can't lorry from zip archives.15:08
KinnisonCurrently we cannot, no15:09
ssam2would that be useful? I got a question from a customer about that the other day15:10
straycatI was going to leave it until we see there's a need for it.15:11
straycatBut if we needed it I can't imagine it would be difficult to add support to lorry.15:11
KinnisonThe trick will be writing something to turn a zip file into a git fast-import stream15:11
KinnisonWe have some perl in lorry for doing that for tarballs15:11
jjardonKinnison: mmm, cant the trove have control of that in some way? Is it not possible to configure trove to not allow to remove tags? with a pre-receive or pre-update trigger maybe? Currently seems baserock is trying to solve the problem of "not desired changues upstream" in 2 different locations: in the trove (not allowing force-pushes) and in definitions15:16
jjardon(making mandatory the use of refs)15:16
Kinnisonthe use of sha1s also means we don't have to refer to the repo to resolve the ref15:18
ssam2that's true and there have been a few solutions proposed in the past15:18
Kinnisonwhich speeds things up15:18
KinnisonUltimately I want to get rid of referring to anything but the committed content of definitions to know how to build the artifact graph15:19
ssam2jjardon: and have been proposed15:21
richard_mawis the gerrit down? I can't reach it currently15:21
pedroalvarezrichard_maw: yes, one second, I was trying one thing15:23
richard_mawaah, it's green!15:24
jjardonssam2: thanks, I somehow missed the September email: Its exactly what I was thinking of!15:24
richard_mawoops, accidentally clicked the submit button, now it's merged it onto the repository on the gerrit15:25
pedroalvarezrichard_maw: how was your merging experience>15:27
richard_mawwell, now we need to work out how it's going to apply that to git.baserock.org15:28
Kinnisonideally we'd make them one and the same thing eventually15:28
Kinnisonbut zuul won't have to use gerrit's merge stuff15:28
jjardonssam2: so, I think your idea is the one that makes more sense. In the end of the day upsrteam repos that remove/move tags are the exception, not the rule15:29
richard_mawI disagree, it often happens in releases15:29
pedroalvarezrichard_maw: just to let you know, your patch is not going to appear in g.b.o :)15:30
richard_mawI know15:30
pedroalvarezjust in case15:30
richard_mawnot unless I push it myself15:30
richard_mawwhich defeats the point of having the review bit really15:31
richard_mawwell, maybe not entirely15:31
KinnisonLooks like we really need the zuul bit now, even if only as the gatekeeper to master15:31
richard_mawas there's still tracking patches going on, but merging is manual15:31
Kinnisonpedroalvarez: Can you change the send address?15:34
Kinnisonpedroalvarez: it's very distracting getting mail from you about things from gerrit :-)15:34
pedroalvarezI can, what should I put?15:34
pedroalvarezI had problems with fake addres15:34
rjekFrom: Gerrit Rietveld <>15:35
jjardonrichard_maw: I didnt say it doesnt happen, but I really cant think the majority of maintainers use the tags incorrectly15:35
rjekpedroalvarez: It's who gerrit is named for:
ssam2richard_maw: in the case that someone's making a release and pushes the same tag a few times, it's unlikely that Baserock would be consuming the release tag any valuable versions of definitions.git straight away15:36
ssam2*in any valuable versions of definitions.git15:36 would be acceptable IMO15:36
Kinnisonone sec15:36
Kinnisonpedroalvarez: give it 60s and will forward to you15:38
pedroalvarezaddress changed15:41
pedroalvareznow you can filter them and throw them to de bin15:41
pedroalvarezI have to stop thinking phonetically15:42
franredpedroalvarez, hahaha or jajaja15:42
*** tlsa [REVts2I8rW@gateway/shell/pepperfish/x-aftcrstyufiffyre] has quit [Ping timeout: 260 seconds]15:48
franredare we interested to update cpython to v2.7.8? this will avoid us to use some monkey-patches done in the pass and probably another errors15:53
franredI can easily make a patch for it15:54
pedroalvareza gerrit patch? :P15:55
franredif I can register my email sure15:56
ssam2franred: there'll be security fixes too, it's a good thing16:02
richard_mawit's probably a good idea that my gerrit patch wasn't automatically applied to definitions.git16:08
richard_mawthere is no /dev/stdin at that point apparently16:08
richard_mawhence the need for the review bot16:08
*** dutch [] has quit [Quit: Quit]16:09
*** tiagogomes [] has quit [Ping timeout: 258 seconds]16:22
franredpython update sent for review!! \o/ thanks pedroalvarez for the configuration and the help!!16:28
pedroalvarezrichard_maw: so is your reverting patch a real patch?16:29
pedroalvarezdo you want to revert the change?16:29
richard_mawthat depends on whether we treat it as having really been applied16:29
pedroalvarezmeh, having 2 differents repositories is confusing me16:30
*** jonathanmaw [] has quit [Quit: Leaving]16:31
franredshould we send patches to the mailing list? or are the people looking at  for review patches?17:07
*** mSher [] has quit [Quit: Leaving]17:19
ssam2I think it's worth notifying the mailing list if you post something to the experimental gerrit instance17:27
franredok, I will do it17:32
*** franred [] has quit [Ping timeout: 260 seconds]17:41
*** ssam2 [] has quit [Read error: Connection reset by peer]17:51
*** De|ta [~arc@] has quit [Quit: leaving]21:24
pedroalvarezSearch results for Zuul are really confusing.. Maybe I was wrong saying that we need a JSP to run it.22:01
rjekThere is no JSP, only Zuul.22:06
*** genii [~quassel@ubuntu/member/genii] has quit [Read error: Connection reset by peer]22:13
pedroalvarezWhich is python22:15
pedroalvarezI cant find instructions about how to configure it 22:16
pedroalvarezMaybe I can install it in the Gerrit host. 22:17
* Kinnison heads to bed after that headache of an evening I feel awful :-(22:28

Generated by 2.15.3 by Marius Gedminas - find it at!